topic Re: Simulate data in Functional Data Explorer in Discussions

Simulate data in Functional Data Explorer

fdaf — Wed, 06 Dec 2023 17:10:53 GMT

I would like to simulate curve data using Direct Functional PCA from Functional Data Explorer.

There is a simulator within Direct Functional PCA which is great!

At a closer look, the simulator is simulating points (one can choose the number of points in advance) on the x axis and not separate curves. I tried different tricks to get separate curves, but with no luck.

Therefore, I am wondering if there is a way to fetch the JSL script code from "Make table" and modify that? In my opinion that would be the easiest as I could also specify equidistant points on x for each generated curve.

Thank you in advance!

Re: Simulate data in Functional Data Explorer

Jed_Campbell — Thu, 07 Dec 2023 16:53:49 GMT

Hi @fdaf, welcome to the forum.

I may not be correctly understanding your question, so please correct me if necessary.

To create a rough version of separate curves, I did the following:

Used the sequencing feature in the Simulator
Deleted duplicate rows in the Time - FPC 5 columns (most of the table)
Created a Checksum formula column summing FPC 1 through FPC 5, and used it as the curve label (this could be more elegant, for sure)
Verified separate curves via graph builder

I attached both the original file and the output table.

Re: Simulate data in Functional Data Explorer

fdaf — Fri, 08 Dec 2023 07:04:38 GMT

Hi Jed

Thank you for a comprehensive answer. It looks great.

However, when I run the Mill DOE file, I get the time as 9,707. Thus, I presume you changed the Time parameter a bit different in order to get equidistant points along X(Time). Is this correct?

@Fetching the Make Table code. Is there some script way around it?

Thank you once more.

Re: Simulate data in Functional Data Explorer

Jed_Campbell — Fri, 08 Dec 2023 15:21:55 GMT

I presume you changed the Time parameter a bit different in order to get equidistant points along X(Time). Is this correct?

I believe the 9.707 value is just the default value in the profiler (I didn't change it). Getting the equidistant points along X was accomplished in the Sequencing section (10 points between what it defaulted to).

@Fetching the Make Table code. Is there some script way around it?

I'm unaware of a straightforward way to do this. You might be able to accomplish something similar by using the Save Summaries feature.